Opencv mlp is not working as the doc pretend it should work.

edit

I am new using Opencv and in particular multilayers perceptron (mlp). I have looked up on the docs and I am a bit confused at some points. Basically I am trying to make a basic neural network that outputs a 1 if the input is -1 and also to output a 1 if the input is 1. I am using CvANN_MLP::SIGMOID_SYM as the activation function of each neuron with parameters α = 1 and β = 1. Thus, according to the doc, the output of the mlp should always be inside [-1; 1]. But I am able to get an output greater than 1 with the following code:

//creation of the multilayer perceptron
Mat layers = cv::Mat(3, 1, CV_32SC1);
layers.row(0) = Scalar(1);
layers.row(1) = Scalar(2);
layers.row(2) = Scalar(1);

CvANN_MLP mlp;
mlp.create(layers, CvANN_MLP::SIGMOID_SYM, 1, 1);

//the training inputs and outputs
Mat trainingData(2, 1, CV_32FC1);
Mat trainingClasses(2, 1, CV_32FC1);

trainingData.at<float>(Point(0, 0)) = 1;
trainingData.at<float>(Point(0, 1)) = -1;

trainingClasses.at<float>(Point(0, 0)) = 1;
trainingClasses.at<float>(Point(0, 1)) = 1;

//the training params
CvANN_MLP_TrainParams params;
CvTermCriteria criteria;
criteria.max_iter = 1;
criteria.epsilon = 0.00001f;
criteria.type = CV_TERMCRIT_ITER | CV_TERMCRIT_EPS;
params.train_method = CvANN_MLP_TrainParams::BACKPROP;
params.bp_dw_scale = 1.0f;
params.bp_moment_scale = 1.0f;
params.term_crit = criteria;

mlp.train(trainingData, trainingClasses, Mat(), Mat(), params);

for (int j(0); j < trainingData.rows; j++)
{
    Mat input = trainingData.row(j);
    Mat output(1, 1, CV_32FC1);

    mlp.predict(input, output);

    cout << output.at<float>(0, 0) << " ";
}
cout << endl;

This code outputs in the console: "1.06762 0.972991".
We can see that the mlp outputs 1.06762 instead of 1 when the input is equal to 1. Am I doing somethink wrong ?